Benchmark RadarAI BENCHMARK PROFILE
SciFigQual-Bench
A benchmark for evaluating scientific figure quality across five dimensions with full-manuscript context, using 6,308 expert-rated images from top CS conferences and a fixed eval1200 test split.
- Released
- 2026-07-29
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Existing image quality assessment methods are unsuitable for scientific figures; this benchmark provides a contextual, multi-dimensional standard for automated evaluation, enabling model comparison.
Motivation
Scientific images are the core elements of presenting experimental conclusions, elaborating system architecture, and supporting comparative arguments in scientific papers.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.